Active contour model

Snakes, also called active contours, are a concept that is used in digital image processing to determine an object contour. In practice, Snake algorithms are mainly used in medical image processing, such as in the diagnosis of ultrasound images. They are used for computer-based object tracking and are invariant to scaling and rotation.

The concept is based on the description of the object contour by a parametric curve. Their shape is corrected as a function of so-called internal and external energies often after a manual initialization. The external energies calculated here from the image content with respect to the position of the contour. Often this is a form of gradient is used ( Gradient Vector Flow). The internal energies are calculated solely from the shape of the contour. By means of a minimizing algorithm, the shape of the contour is calculated, in which the sum of the energy reaches a minimum. Instead of the minimization algorithm, the shape of the snake is often changed and then considered that form as a result in which the sum of the energies is a minimum, practically by trial and error.

The first report of Snakes is the work of Kass, Witkin and Terzopoulos. Since then, followed by numerous other articles that suggest, inter alia, new types of energy functionals and new approaches to minimize the total energy.